Skip to main content

San Francisco plastic-bag ban associated with 46% increase in foodborne illness deaths — Not!

Tomás Aragón, Clinical Faculty, School of Public Health | February 13, 2013

In my role as Health Officer of San Francisco I received a flurry of concerned calls about a research study that claimed that the 2007 San Francisco ban on plastic bag resulted in an immediate, very large increase in foodborne illnesses and deaths. From their conclusions:

“We examine deaths and emergency room admissions related to these bacteria in the wake of the San Francisco ban. We find that both deaths and ER visits spiked as soon as the ban went into effect. Relative to other counties, deaths in San Francisco increase by almost 50 percent, and ER visits increase by a comparable amount.”

Here is the introduction to a memo I composed in response to this study:

“This memo is to respond to a recent unpublished research paper concluding that the San Francisco ban on plastic bags has led to an increase in bacterial foodborne illnesses and deaths. This paper is from Jonathan Klick and Joshua D. Wright from the University of Pennsylvania Institute for Law & Economics. Based on our review of this paper, and our disease surveillance and death registry data, the Klick & Wright’s conclusion that San Francisco’s policy of banning of plastic bags has caused a significant increase in gastrointestinal bacterial infections and a “46 percent increase in the deaths from foodborne illnesses” is not warranted.”

My full memo is here:


The Klick study I read is here:


Reading their article and responding to public concerns reminded me of several things:

  • Law professors and epidemiologists use difference study designs to infer causality.
  • Research studies with alarming conclusions can alarm the public  —  so be cautious.
  • Reach out and collaborate with experts in other disciplines — the science is better.
  • Ecological study designs should not be the basis of exposure-disease causal inference when better methods are available. If used, stress the limitations and assumptions.
  • Assessing the health impacts of policies is called Health Impact Assessments (HIAs). Consider using the Institute of Medicine HIA Framework.

Comments to “San Francisco plastic-bag ban associated with 46% increase in foodborne illness deaths — Not!

  1. What is your response to the 2010 University of Arizona / Loma Linda assessment that found that only approximately 3% of people cleaned their reusable bags on a regular basis, and found coliform and HPC bacteria in more than half?

  2. Where there’s smoke (admittedly dirty reusable bags in the public domain) a public health fire of some sort would seem to be expected.

  3. Dr. Aragon’s memo leaves me unconvinced. The strongest evidence in the Klick study is that the increases in ER visits occurred in the other localities with bans on plastic bags. Dr. Aragon ignored this.

    He criticized the economists for using the methods of economics and not that of epidemiology. Well, they aren’t doing epidemiology. Epidemiologists are usually establishing causal relationships so they can treat diseases. Their standards of causality are much higher. Dr. Aragon also criticizes the study in that it did not follow individual cases. It is not necessary, in economics or epidemiology, for the person whose behavior causes the disease, to be the one in whom the disease manifests. In other words, I don’t wash the bag, my aging mother contracts disease. This causal relationship seems to escape Dr. Aragon’s.

    What the economists did, almost as an aside, was to show what a trivial effect such municipal bans have on the environment. Anti-environmentalists point to such policies to discredit the environmental movement in general. When you talk about a real problem like global warming, you lose credibility if you also backed bans on plastic bags.

    • 1. ER visits are incomplete data, we look at all the data for a jurisdiction.

      2. See course on causal inference from Northwestern Law:
      Workshop on Research Design for Causal Inference:
      – Main Workshop: June 24-28, 2013
      – Advanced Workshop: August 12-14, 2013

      We use the same science.

      3. “It is not necessary, in economics or epidemiology, for the person whose behavior causes the disease, to be the one in whom the disease manifests. In other words, I don’t wash the bag, my aging mother contracts disease.” Correct! Your aging mother with DISEASE was EXPOSED to a contaminated bag. You could only know that by questioning individuals.

      4. As I wrote in the memo, ecological studies are good for generating hypothesis not testing them.

      5. Given their study design and inaccuracy with the death data, their conclusions were not warranted.

  4. I must strongly agree and emphasize that an associated observation is insufficient evidence for causality.

  5. Thanks for this response. I agree it is unfortunate that so many news outlets picked up this story, and editors came up with such alarming headlines, without getting an opinion like this one.

    Sure, this is an important topic to bring up and study. We should be encouraging everyone to wash their produce quite thoroughly as well. But I was disturbed by the causality inference.

    When these bans are implemented, stores, local governments and NGO’s hand out thousands of free, reusable bags. The reusable bags being used in those first few months are probably the cleanest bags ever.

    Paper bags were (and still are) still available. And, as you point out, who is to say that ANY of those people hospitalized or those who died even used a reusable bag??

  6. Typically epidemiologists look at much more detailed data and would be much more likely to determine a causal relationship. There are privacy issues involved and review standards required to get access to highly confidential data and patients. As one can see in the news, public health researchers will even track down what people ate to trace an outbreak.

    First, I doubt one can successfully sue the state or city for passing a law even if it caused more deaths; if so all states would be broke. Second, a decent judge after hearing from an epidemiologist would probably not even allow this study.

    I’m neither a law professor or a health researcher.

  7. It would be great to have some plain language explanations here, rather than just linking to official documents. Isn’t the format of a blog most useful for introducing ideas to people who aren’t yet well-informed on a subject?

  8. Thank you so much for this response to the flurry of media which has been growing since this info came out. Unfortunately many consumers do not have time to do their own research and thus, believe without question.

    Amazing all of these places have bag bans and this has never been brought up before. Who paid for the research?

    Off to promote this article!

    Shane (of

    • “Who paid for the research?”…. Hmmm….I’m asking myself exactly the same question!! So sad!!

  9. I wonder if you can put on your blog the actual numbers for the E. Coli report history, as the relatively low frequency is crowded near the bottom of the graph.

    Also, I was curious about “selected laboratory” results that are positive for E. Coli must be reported; would any listed as cases “E. Coli” in the emergency room data not be reported? Is the criteria for which results are reported based on severity of disease or the specificity of the testing method used?

    • We are all colonized with E. coli, so our stool cultures will always be positive with “coliforms.” With diarrhea, the labs look for Shiga-toxin producing E. coli (STEC) that can cause bloody diarrhea and kidney failure in the young. The data used for my response are publicly available in annual reports here:

      There are other strains of E. coli (enterotoxigenic, enteropathogenic, enteroaggregative, enteroinvasive, etc.) that are not commonly tested by labs. The primary focus is detecting STEC. CDC has excellent info here:

      • Thanks for the sources. They pretty much show that any risk from E Coli being spread due to plastic bag bans is below the limit of detection in San Francisco. It is too bad that the authors were not more careful before shouting “fire”.

  10. Thank you! I’m guessing they had some funding from the plastics industry. I’m disappointed that mainstream media like the Huffington Post published articles about this non-peer-reviewed, questionable study. I hope they publish your results as well.

      • The most common cause of these type of errors is an ideological bias that prevented a more careful questioning of the data and its usage. Studies that support an author’s thesis are published; those that don’t get more careful scrutiny.

  11. You stated that “Law professors and epidemiologists use difference study designs to infer causality.” But which study designs will be used as causality in civil suits against the State of California?

Comments are closed.